Multi-scale computational homogenization is a method adopted to provide a numerical solution to the nonlinear response of heterogeneous problems, potentially involving discontinuities. Several solutions have been proposed, adopting either traditional numerical schemes like the finite element method, or machine learning techniques [1]. Therefore, relevant studies use neural network metamodels to feed the upper-level continuum with a homogenized constitutive law based on the solution of Representative Volume Elements (RVE) for different loading cases. A class of advanced machine learning methods introduce in the loss function the physics of the structural analysis problem, mainly the equilibrium equations and the constitutive description. These models, which are widely known as Physics-Informed Neural Networks (PINNs) [2], use feed-forward neural networks, trained by the backpropagation algorithm and automatic differentiation, so that the residuals include governing differential equations and constitutive descriptions. Despite the wide use of PINNs in several computational mechanics applications [3], less efforts focus on applying PINN solutions to heterogeneous materials, where discontinuities between the constitutive materials and corresponding interfaces, potentially appear. The reason for this is the fact that PINNs cannot provide accurate solutions when discontinuities are present. Therefore, the aim of this work, is to propose a solution addressing the issue of using PINNs in solving structural problems for heterogeneous materials. To achieve this goal, a multi-surrogate PINN scheme is developed, consisting of different subnetworks simulating each of the RVE material domains. A unilateral contact interface providing the interaction between the constituent materials of the RVE is also considered by the proposed PINN. In total, the study provides a methodology of using PINNs in the context of multi-scale computational homogenization, by applying periodic boundary conditions and determining effective stress and stiffness of the microstructure.